0:01

Welcome to Practical AI,

0:03

the podcast that makes artificial

0:05

intelligence practical, productive, and accessible

0:07

to all. If you like

0:09

this show, you will love

0:11

the change log. It's news

0:13

on Mondays, deep technical interviews

0:15

on Wednesdays, and on Fridays,

0:17

an awesome talk show for

0:19

your weekend enjoyment. Find us

0:21

by searching for The Change

0:23

Log, wherever you get your

0:26

podcasts. Thanks to our partners

0:28

at fly.io. Launch your AI

0:30

apps in five minutes or

0:32

less. Learn how at fly.io.

0:44

Welcome to another episode of

0:46

the practical AI podcast. This

0:48

is Daniel Whitenack. I'm CEO

0:50

at Prediction Guard, and I'm

0:52

joined as always by my

0:54

co-host Chris Benson, who is

0:56

a principal AI research engineer

0:58

at Lockheed Martin. How you

1:00

doing, Chris? Oh, I'm feeling

1:02

pretty chipper today. It's a

1:04

good day to talk about

1:06

AI. Yeah, yeah, I feel

1:08

quite quite chipper as well,

1:10

especially as we've got our

1:12

guest today, Scott Meyer with

1:14

us, who's founder and CEO at

1:16

CHIP, which is, you can find

1:19

at CHIP.a.i. I believe is the

1:21

link, but yeah, CHIP is awesome.

1:23

Also, Scott is. is along a

1:26

good friend because he's a fellow

1:28

member of the Silicon Prairie, not

1:30

living on the coast, but out

1:33

here in the middle somewhere where

1:35

AI is really blossoming if you

1:37

didn't know? It is. And it

1:40

gives an unfair advantage for those

1:42

of us in non-metro areas, you

1:44

know, like the ability to leverage AI

1:46

to have the power of 10 people

1:48

in a place that doesn't have enough

1:50

people to do the job. It's perfect.

1:53

It's a perfect solution. So it's great

1:55

to be here live from Fargo, just

1:57

like the movie. It's fantastic to see

1:59

you all. be heard by all of

2:01

you listening. Yeah, yeah, Scott, well, we'll

2:03

get into all the cool stuff, you

2:05

know, you're doing with Chip and some

2:08

of the things you've learned through that,

2:10

but I'm wondering if, you know, you

2:12

work in the space of, I guess

2:14

we might put it like low code,

2:16

no code, AI assistant builders. So for

2:18

maybe audience members that aren't as familiar

2:21

with that space or maybe they're just

2:23

kind of wondering what's out there. you

2:25

know, as of as of today, could

2:27

you paint a little bit of a

2:29

picture for us for kind of what

2:31

sorts of tools are out there? And

2:34

then maybe that would kind of motivate

2:36

some of the unique things that you

2:38

thought should be out there but weren't,

2:40

which would maybe kind of highlight some

2:42

of the things you're doing with Chip.

2:45

Yeah, now it's great to be here.

2:47

I think. The staff that blows my

2:49

mind is that almost 50% of Americans

2:51

use AI every week, but 7% of

2:53

businesses use AI, which is obviously a

2:55

lie, because 50% of Americans are using

2:58

AI every week and they work at

3:00

those companies. So what's happening is the

3:02

businesses, aren't, what's happening is the businesses

3:04

aren't, they have no idea what's going

3:06

on. It's like the early days of

3:08

cell phones, when everyone would come to

3:11

work with their own cell phone, their

3:13

own laptop, do whatever they wanted to,

3:15

do it. I think the risk right

3:17

now is that, and the opportunity, is

3:19

those who are willing to have agency

3:22

and try stuff have unfair advantage, right?

3:24

So I can go do my work

3:26

with AI, and if my colleagues don't

3:28

know, and I don't have a culture

3:30

of sharing, like all of a sudden

3:32

I'm a super human, the number one

3:35

thing I tell businesses when I meet

3:37

with them is you should have a

3:39

lunch and learn once a month and

3:41

just have people say what they're doing.

3:43

Because just that horizontal sharing. of AI

3:45

practices and ideas is all you need

3:48

to build a culture of acceptance. And

3:50

what makes AI so unique is it's

3:52

not top down. It's not the CIO

3:54

or CTO saying, I bought this thing,

3:56

you guys all go use it. It's

3:59

each individual figuring out how they can

4:01

use it for their specific tasks. And

4:03

what I've seen is admin assistance, you

4:05

know, marketers, interns, right, they're all gonna

4:07

use it differently and often even know

4:09

better how to use it because they're

4:12

the ones doing the tasks. And that

4:14

kind of motivated what we built with

4:16

CHIP, which is how do we just

4:18

make AI as easy as possible to

4:20

use? Our, you know, kind of our

4:23

motto is AI for all. And I

4:25

think I've spent most of my professional

4:27

career working on. bridging a digital divide

4:29

because maybe like you, you know, people

4:31

that work and live alongside me in

4:33

Fargo aren't always taking advantage of the

4:36

latest technology, right? And so I kind

4:38

of feel like it's both a passion

4:40

and mission to bring what's happening and

4:42

make it accessible to those around me.

4:44

In 2009, I started my first company

4:46

and I was trying to tell businesses

4:49

there's this thing called social media they

4:51

should use, right, before there are Facebook

4:53

pages and Facebook ads and it feels

4:55

like that to me again almost 20

4:57

years later where it's like. this amazing

5:00

power is right here and the best

5:02

time to start learning is now. And

5:04

with tools like Chip and others that

5:06

we can talk about, it's actually better

5:08

now than ever for people who aren't

5:10

technical because it's not about technical ability,

5:13

it's about knowledge and agency. And I

5:15

think we all have that. So happy

5:17

to give a landscape. I think that

5:19

already went off track from your question,

5:21

but hopefully that gives you a starting

5:23

point. No, that's awesome. What would you

5:26

say are kind of some of those

5:28

things that might make AI hard to

5:30

use. And here, you know, mostly we're

5:32

talking, of course, we've talked about a

5:34

lot of things in the show, but

5:37

mostly we're talking about kind of what

5:39

typical people would consider AI now, which

5:41

would be kind of generative AI language

5:43

models, maybe vision models, etc. So like,

5:45

what can make those difficult to use

5:47

or how might people get disillusioned as

5:50

they're exploring the technology? I'll say almost.

5:52

Almost every excuse people have not to

5:54

use AI tools is fear. They are

5:56

scared of a blank page. And this

5:58

is the same with technology for 20

6:00

years. I taught entrepreneurship. I started entrepreneurship

6:03

centers and all these students with a.

6:05

ideas and you know what 90% of

6:07

them didn't do anything because they had

6:09

to actually go do something right and

6:11

it's like you just have to start

6:14

and I'm convinced the biggest challenge in

6:16

AI is change management it's just getting

6:18

people to to start and I think

6:20

this happened when Google first came out

6:22

you know it's a blank screen blank

6:24

prompt window, like what do I say

6:27

when I can say anything? It's actually

6:29

quite intimidating. And so that's the challenge

6:31

I think with AI is like anything's

6:33

possible. So where do you start? I

6:35

tell everybody the best place to start

6:38

is to create your digital protege. Like

6:40

just tell AI what you do and

6:42

have it help you do those things.

6:44

AI is great at what you hate.

6:46

And so find those things that you

6:48

hate doing or that take a lot

6:51

of time and start there. You've maybe

6:53

seen that quote. dishes and laundry so

6:55

I can do more art and music

6:57

not AI to do art and music

6:59

so I can do more dishes and

7:01

laundry right so I think we all

7:04

have dishes and laundry in our day-to-day

7:06

life and so let's use AI there

7:08

first because that'll be the you'll get

7:10

more motivated to do fewer financial analyses

7:12

or fewer I don't know copy editing

7:15

because that's kind of annoying than you

7:17

would like making music because maybe that's

7:19

fun for you right so start with

7:21

things that you don't like one thing

7:23

I find fascinating about research on AI

7:25

is actually having knowledge makes you better

7:28

positioned to use AI. I think about

7:30

AI as like the rebirth of the

7:32

Renaissance person. It's like if I want

7:34

to create a picture on AI that

7:36

looks like Picasso, but I don't know

7:38

Picasso's name, it's really hard to describe

7:41

that, right? If I want to make

7:43

a blueprint of a Georgian architecture building,

7:45

like how do I explain that if

7:47

I don't know what Georgian architecture is?

7:49

And so whatever area you live in

7:52

or work in or care about. You

7:54

have, like, expertise, right? You can talk

7:56

about it all day. And that's a

7:58

great place to start with AI, because

8:00

you can go say those words, like,

8:02

give me, I don't know, a hierarchy

8:05

of Pokemon characters, and you can name

8:07

all the things and have it rank

8:09

order it. Like, I have no idea

8:11

what I. I would say for that

8:13

right, but I can talk all day

8:15

about saunas and have the AI help

8:18

me improve my sauna find new water

8:20

buckets look at you know different ratios

8:22

of time in the sauna like because

8:24

I care about that so find some

8:26

things that you know about that you're

8:29

passionate about and start asking AI about

8:31

it so you can go deeper I

8:33

love it. I'm curious quick follow-up on

8:35

that you know because you raised a

8:37

point that I hadn't really thought about

8:39

but I've observed it many many many

8:42

times and I've with, I see people

8:44

who are totally comfortable getting on the

8:46

search engine of their choice and searching

8:48

topics and they've been doing that for

8:50

years, but as soon as they pull

8:53

up, you know, a chat with a

8:55

given model, they're really struggling with that.

8:57

And they're really, that's what. I'm just

8:59

curious as you've clearly thought about this

9:01

quite a lot. What is the difference?

9:03

And why are people so easy to

9:06

go to search and yet struggling with

9:08

that model that has the same text

9:10

box in front of it? Part of

9:12

its exposure, right, just history, but I

9:14

also think there's something quite vulnerable about

9:16

AI where it's really a two-way conversation.

9:19

Search engine is, you know, very much

9:21

like the old card catalogs. You know,

9:23

I remember my... First year of elementary

9:25

school, I learned card catalog and then

9:27

the next year was told never have

9:30

to touch that again. But it's the

9:32

same, that worked the same, right? I'm

9:34

just going to go find something. But

9:36

with AI, it's probing back and forth

9:38

and actually you can get, you can

9:40

get pushback and it kind of identifies

9:43

how you're thinking about things. So I

9:45

think there's some vulnerability around that and

9:47

plenty of like blank page problem of

9:49

just not knowing where to know where

9:51

to start, framework to get started is

9:53

what I call the ripe framework so

9:56

RIPE and it's just a way of

9:58

like four sentences to put into AI

10:00

to get good answers which is the

10:02

role so like you are an expert

10:04

I don't know copy editor the instruction

10:07

like read through my paper and improve

10:09

it parameters so make sure it's very

10:11

concise and don't repeat a lot of

10:13

the same points and examples like here's

10:15

a paper I wrote before that shows

10:17

my kind of tone. You know if

10:20

you just do those four things a

10:22

role instruction parameter example like you're gonna

10:24

get awesome output that's personalized and much

10:26

more effective and less robotic than just

10:28

going there and saying write me a

10:30

paper. Yeah I've had this kind of

10:33

hypothesis I guess going around in my

10:35

mind I'm curious Scott on your on

10:37

your take on this. Because you've seen

10:39

a lot of people now, you're always

10:41

interacting with people on this quarter, wherever,

10:44

you know, trying to get their assistance

10:46

to do this or that. What have

10:48

you found to be kind of the

10:50

qualities that make up someone who is

10:52

just really proficient at kind of honing

10:54

in the the instructions, the... data integration,

10:57

the configuration of AI systems. My hypothesis

10:59

is sort of this is almost like

11:01

a, I think if we took a

11:03

bunch of hostage negotiators and had them

11:05

log in to AI systems to try

11:08

to, you know, either get them to

11:10

do things that they wanted them to

11:12

do or to jailbreak them, I think

11:14

they would be like amazing at this,

11:16

because a lot of times it seems

11:18

to me. you know, not that I

11:21

feel in danger physically or something, but

11:23

it's like people can get disillusioned with

11:25

this. It's like not quite what I

11:27

want. How do I get you to

11:29

do what I want you to do?

11:31

How do I like warm you up

11:34

to this idea? So yeah, I'm curious

11:36

on the qualities that you've seen in

11:38

terms of people that have become good

11:40

at configuring these systems, prompting, understanding how

11:42

to... you know, pull in integrations or

11:45

when and where to do that. Any

11:47

thoughts? Yeah. I mean, people who are

11:49

great at this are kindergarten teachers or

11:51

parents of three-year-olds. Maybe also hostage negotiation.

11:53

Also, it's basically the same thing. There's

11:55

some similarity there maybe? Yeah, I mean,

11:58

think about talking, I mean, people, I

12:00

say an intern, but that's even too,

12:02

too experienced. Think about talking to my

12:04

three-year-old Sebastian. If I tell him three

12:06

things to go do in order, there's

12:08

no way he's gonna get all three

12:11

of them done. Right? Like, go to

12:13

the bathroom, pick out some shoes, grab

12:15

your snack, go to the car. Like,

12:17

that's not happening. I have to be

12:19

like, go to be like, go to

12:22

the bathroom. Good, now this, right? And

12:24

now this. It's very step-by-step. And I

12:26

think what's interesting is there's two models

12:28

or two types of models emerging in

12:30

AI, and you guys maybe have your

12:32

own language for this, but I think

12:35

about linear models, like 4.0, Claude Sonnet

12:37

35, and we have reasoning models now,

12:39

like O3, Deep Seek, and now Sonnet

12:41

37. And it's like, the reasoning models

12:43

actually, that's like talking to an intern,

12:45

who you can give a ton of

12:48

stuff, and you just let it just

12:50

let it go. But if you're doing

12:52

a linear model, that's very much need

12:54

to do that step by step. First

12:56

do this, then do this, then do

12:59

this. Because the biggest... I think frustration

13:01

people have is that AI too quickly

13:03

tries to get to an answer before

13:05

it has all the details and things

13:07

get lost. And so with chip, you

13:09

know, you can prompt your AI tool

13:12

and then anyone can use it. And

13:14

so what we found is like flipping

13:16

the relationship is really powerful where the

13:18

AI prompts you to get what it

13:20

needs and then gives an answer. So

13:23

you can even, you know, on chip,

13:25

you can build this in so you

13:27

don't have to. type it every time

13:29

but on any AI tool you might

13:31

say like before you write the paper

13:33

before you create the you know strategy

13:36

before you create the I don't know

13:38

the press release make sure to ask

13:40

me these three things right and force

13:42

it to get all of that information

13:44

step by step just like you do

13:46

with a three-year-old and then you write

13:49

the paper and then you do the

13:51

thing right so I think that's really

13:53

fascinating though seeing that divergence with reasoning

13:55

which is like don't go step by

13:57

step. Just give all the context and

14:00

it's going to work through it on

14:02

its own versus the three-year-old linear that

14:04

needs that guidance. So yeah, I think,

14:06

and the end of the day, hostage

14:08

negotiator and parents, you've got. this. Well

14:10

friends today's ever-changing AI landscape means your

14:13

data demands more than... Well friends today's

14:15

ever-changing AI landscape means your data demands

14:17

more than the narrow applications and single

14:19

model solutions offer. Domo's AI and data

14:21

products platform is a more robust all-in-one

14:23

solution for your data. It's not just

14:26

ambitious, it's practical and adaptable. So your

14:28

business can meet those new challenges with

14:30

ease. With Domo you and your team

14:32

can channel AI and data into innovative

14:34

uses that deliver measurable impact, and their

14:36

all-in-one platform brings you trustworthy AI results

14:38

without having to overhaul your entire data

14:41

infrastructure, secure AI agents that connect, prepare,

14:43

and automate your workflows, helping you and

14:45

your team to gain insights, receive alerts,

14:47

and act with ease through guided apps

14:49

tailored to your role, and the flexibility

14:51

to choose which AI models you want

14:53

to use. Domo goes beyond productivity. It's

14:56

designed to transform your processes, helping you

14:58

make smarter and faster decisions that drive

15:00

real growth. All power by Domo's trust,

15:02

flexibility, and their years of expertise in

15:04

data and AI innovation. Data is hard.

15:06

Domo is easy. Make smarter decisions and

15:09

unlock your data's full potential with Domo.

15:11

Learn more today at AI. So Scott,

15:13

maybe we'll come back to kind of

15:15

the tooling itself. Could you maybe kind

15:17

of circle back and describe some of

15:19

the Maybe people aren't familiar with some

15:21

of the kinds of tools that are

15:24

out there, especially, you know, maybe there's

15:26

programmers that have interacted with APIs that

15:28

are listening to the show. Maybe there

15:30

are people that have explored one tool

15:32

or another. Maybe there's people that haven't

15:34

explored anything yet. So could you maybe

15:37

just help us kind of form a

15:39

mental model for the kinds of AI

15:41

tools that are out there? And then

15:43

maybe that would lead into. Yeah, a

15:45

discussion about kind of some of the

15:47

things that that were really on your

15:49

mind in terms of needs that weren't

15:52

being addressed in that ecosystem. Yeah, I

15:54

mean, if you want to think of

15:56

like a simple two by two matrix,

15:58

I think there's a really clear like

16:00

vertical versus horizontal and like closed versus

16:02

open dichotomy. So you can think about

16:04

horizontal tools doing a lot of things

16:07

across modes, right? So jetch EPT can

16:09

write, it can create images, it can

16:11

code. It's really good at all of

16:13

those, but if you want to just

16:15

make images, like mid-journey is probably better,

16:17

right? It's a vertical image generation tool,

16:20

or PECA is really good at video

16:22

generation, which some of the general horizontal

16:24

tools aren't as good at. And, you

16:26

know, my sense is like horizontal is

16:28

going to win, but there's always going

16:30

to be a need for people who

16:32

want the... Maserati of AI, right? If

16:35

you're only doing code, like you're going

16:37

to probably be in cursor going deep

16:39

into like using these tools, whereas someone

16:41

like me, I'm going to do the

16:43

vibe coding where I can use a

16:45

tool like lovable or bolt and just

16:47

try stuff or replet, right? So, I

16:50

think horizontal, vertical, and then I think,

16:52

you know, kind of open close. So,

16:54

there are tools that let you, you

16:56

know, use it on their platform and

16:58

you don't necessarily know what's happening. So

17:00

that would be obviously like Chatch EPT

17:03

or Claude, you can't change the kind

17:05

of rules underpinning it. Also, you can't

17:07

change the kind of rules underpinning it.

17:09

Also, you have to go to their

17:11

website, you can't brand it, you know,

17:13

you know, we want to bring the

17:15

power of AI tools like. Chatuchy BT

17:18

and Claude to your website, add privacy

17:20

so the file stay locally, add your

17:22

own branding, you can see the chat

17:24

log. So just a lot more control,

17:26

obviously like prediction guard, same thing, right,

17:28

where you can bring AI into your

17:30

own cloud. So a little bit more

17:33

work, obviously, with an open tool, or

17:35

you more power, also you get more

17:37

options. So I think that's kind of

17:39

like the lay of the land. And

17:41

I think it's just like when you

17:43

look at the internet broadly, like it's

17:46

text because that was easy to send

17:48

across wires and then music because M3-3s

17:50

were smaller than video and then video

17:52

see the same thing with AI right

17:54

where it started with text and code

17:56

because that's text heavy starting to get

17:58

pretty good images now video is still

18:01

coming not quite there yet but getting

18:03

better every day so I kind of

18:05

see that evolution happening yeah and I

18:07

think what Maybe it's a surprise is

18:09

that people thought the value was in

18:11

the large language models. And I think

18:14

what's become really clear the last month

18:16

or two is it's actually gonna be

18:18

in the customer relationship and making the

18:20

stuff easier to use. Deep Seek is

18:22

the model that came out of China

18:24

a couple weeks ago. And you know,

18:26

if I look at what CHIP's AP,

18:29

like cost per API call is, it's

18:31

gone down 90% in 18 months, right?

18:33

So just think about the value of

18:35

these large language models becoming more commoditized.

18:37

And then what. What is what people

18:39

are signing up for is like the

18:41

experience of signing up and creating. So

18:44

I can go to Replet and say

18:46

I want an app that, you know,

18:48

is tracking my to-does and get it

18:50

in a few minutes. It's all on

18:52

top of the same power, right? It's

18:54

all on top of chat cheap, you

18:57

can use any model underneath, but it's

18:59

that end user experience, which maybe isn't

19:01

so different than the web, right? There's.

19:03

protocols underneath, but you still use the

19:05

browser that you like or the web

19:07

app you like because of how it

19:09

works, not necessarily that it uses FTP

19:12

versus something else. Could you talk a

19:14

little bit more about that end-user experience,

19:16

both the good and the bad? Because

19:18

I think, you know, kind of going

19:20

back to what we were talking about

19:22

before, it's one of those barriers and

19:24

you know there's a set of people

19:27

that are totally bought in across a

19:29

whole bunch of different industries but there's

19:31

also a very large second of the

19:33

population that still really hasn't engaged. You

19:35

know they're hearing about it every day

19:37

in the news and everything but they're

19:40

just intimidated and haven't done it. So

19:42

could you talk a little bit about

19:44

the landscape of being on both sides

19:46

of that barrier for different people? I

19:48

mean the biggest increase in use that

19:50

we see with AI is putting it

19:52

where people already are so they don't

19:55

have to learn a new interface, right?

19:57

So if they can engage with AI

19:59

via a slack channel or via WhatsApp

20:01

or via text message, like way easier,

20:03

right? And so I think it's fascinating

20:05

to see. There's a lot of amazing

20:08

UIs out there. but it's still like

20:10

getting people there. It seems like time

20:12

to value is really important with the

20:14

tools. So like the faster you can

20:16

show somebody in outcome. And that's I

20:18

think where a lot of the new

20:20

kind of text to app tools, like

20:23

Loveable and Bolt are really exciting for

20:25

people because they can get something quick,

20:27

which makes sense. I think that's kind

20:29

of like how all UI is is

20:31

like how do you get someone to

20:33

the value quickest. I actually think like.

20:35

the default UI we are accustomed to

20:38

with chat TVT is not great. You

20:40

know, like for someone to come in

20:42

there and use, you know, it's interesting

20:44

that. JetGPD was a research project. It

20:46

was not supposed to be a consumer

20:48

app, and it just became that on

20:51

accident. And so I think there's a

20:53

lot of improvements to the UI to

20:55

come to make it easier for people

20:57

to use. And you see those already

20:59

coming into play where there's pre-built ideas,

21:01

auto fill, you know, connect to data

21:03

sources. You know, the most common way

21:06

people use chip is by duplicating it

21:08

in existing app, right? So it's like

21:10

solving that blank page problem is really

21:12

important, I think. to motion is key.

21:14

Yeah, I'm intrigued. You made me think

21:16

of something. So like for those that

21:18

haven't seen Chip and what Scott and

21:21

Team are building, you can go in

21:23

and create individual assistance that, as Scott

21:25

mentioned. and you can kind of control

21:27

and configure, make the way you want,

21:29

connect the data sources you want. And

21:31

often I think in my conversations in

21:34

the past with Scott, I've heard him

21:36

talk about how people are creating, sort

21:38

of proliferating these, right? You create one

21:40

to do this and like, one to

21:42

do that and you clone this one

21:44

to do that because it's not quite

21:46

that, which is a different, it's a

21:49

different paradigm than. this sort of like

21:51

here's a chat interface this chat interface

21:53

is going to do everything that we

21:55

that we want it to do could

21:57

you talk about that that element of

21:59

it a little bit and what you've

22:01

seen there because I I also see

22:04

this on the business side like when

22:06

we engage customers the kind of tendency

22:08

it seems from my perspective is to

22:10

say hey how are we going to

22:12

build like our internal AI? Right. And

22:14

get it to do all the things

22:17

that we want it to do. But

22:19

it's like a single, in their mind,

22:21

it's a single thing, right? It's like

22:23

this is our tool and it's going

22:25

to be the tool to sort of

22:27

rule them all. They're thinking very singularly

22:29

in that way, which definitely does not

22:32

seem to be kind of how people

22:34

are engaging in the way they're building

22:36

assistance in your tooling. Any thoughts there?

22:38

I mean, I think the high level

22:40

thought is the concept of software is

22:42

getting turned on its head where. software

22:45

is now an individual sport, not a

22:47

team sport. You know, you think about

22:49

if you're the CTO even a few

22:51

years ago, it's like, I have to

22:53

do a lot of research by the

22:55

right thing because everyone's going to use

22:57

this. It has to fit the most

23:00

use cases. We have to squeeze everything

23:02

we can into one thing. And now

23:04

it's flipped where every single person can

23:06

build custom software within, you know, we

23:08

say 60 seconds, right? So. You would

23:10

never build software to, I don't know,

23:12

write a better introduction paragraph to a

23:15

grant, but now like someone on chip

23:17

will go build an app that just

23:19

does introductory paragraphs for grant applications, because

23:21

it takes 60 seconds and it saves

23:23

them three minutes every single time. and

23:25

they do 10 a day, and so

23:28

it's 30 minutes. And, you know, we're

23:30

seeing the average admin person saving 60

23:32

minutes a day on ship, going from

23:34

90 minutes to 30 minutes on admin

23:36

work, because they're building specific apps for

23:38

their specific tools. So, you know, today

23:40

I was looking at one that was

23:43

getting IRS status from the IRS website,

23:45

right, and putting it on to a

23:47

spreadsheet. And it's like. Nobody is going

23:49

to go build a sass tool that

23:51

just does that because the market is,

23:53

you know, maybe a hundred people or

23:55

something. But with AI, you can. And

23:58

so there's definitely no need to have

24:00

this like laborious top-down purchase cycle when

24:02

you can say, just try it. Like,

24:04

does this solve one problem? Two problems?

24:06

Five problems? Ten problems? Ten problems? Great.

24:08

Imagine the power of every single person

24:11

in your org being a web developer

24:13

or a coder. Like that's what it

24:15

is now, right? And so now we

24:17

don't have to bother our IT people

24:19

or our developers. They can go do

24:21

like the hard stuff integrating with like

24:23

with antiquated systems, right? Like getting our

24:26

billing to talk to our web to

24:28

talk to this. But for my job,

24:30

I just have a file and I

24:32

need to get something done and like,

24:34

I'm not going to bother our developer,

24:36

but I'm going to be my own

24:38

developer. I don't know, that's a total

24:41

flip, right? Or now we're not making

24:43

decisions for the org, we're making decisions

24:45

for Scott, and I can just build

24:47

it myself, so. the only limiter again

24:49

is is agency like just go you

24:51

have to go do it most people

24:54

still won't even though the tools right

24:56

there but if they can at least

24:58

try once it's not as hard as

25:00

they might think so it's a fascinating

25:02

point you're making there with it but

25:04

it does change that even though you're

25:06

talking about flipping the model over you

25:09

know from kind of catering to the

25:11

the business as a whole to being

25:13

able to cater it to each individual

25:15

contributor in the business by doing that

25:17

I'm curious, you know, that opens up

25:19

a lot of possibilities for how you

25:22

might run the business going forward. Do

25:24

you have any thoughts on like what

25:26

that does to the business if assuming

25:28

what's in a hypothetical world that you

25:30

could get your entire workforce to engage

25:32

in that way? What do you think

25:34

that does? for a business and how

25:37

might, if you were the CEO of

25:39

a business, how might you operate in

25:41

such a way to change that? If

25:43

you were just everyone's empowered with AI

25:45

agents that they can make in 60

25:47

seconds. What does that do for them?

25:49

Yeah, and this is what chips are

25:52

trying to build. This is really my

25:54

Arizona like where we think work is

25:56

going is. We need an umbrella of

25:58

safety so that our employees can do

26:00

whatever they want without feeling like they're

26:02

going to break something. Like right now

26:05

the fear of messing up is greater

26:07

than the fear of missing out. And

26:09

so we need to get rid of

26:11

that fear of messing up. So I

26:13

always say, you know, like the foam

26:15

moo is greater than the foam hole.

26:17

Like we've got to get rid of

26:20

the foam moo because people aren't taking

26:22

action because they're scared. And so. I

26:24

think if I'm a company, what I'm

26:26

doing is I have my 5 to

26:28

10 core apps. This is how we

26:30

work. When you start at Scott Inc.,

26:32

you're going to go through the onboarding

26:35

chatbot. You're going to get the content

26:37

creator that writes everything in our voice.

26:39

You're going to get the data analysis

26:41

that's going to analyze the spreadsheets in

26:43

the same way. So these are the

26:45

apps everybody uses. This is company standard.

26:48

This is getting the laptop with pre-built

26:50

software. And then underneath that now, you

26:52

can duplicate or build your own to

26:54

how you work, right? So you have

26:56

this layer of company-wide apps, and then

26:58

I have my Scott apps, and maybe

27:00

they're only visible to me. And a

27:03

lot of times I might even cross

27:05

personal and professional, potentially, right? Where it's

27:07

like, here's my workout schedule and my

27:09

agenda builder for work and my, I

27:11

don't know, grant writer tool. But since

27:13

it's underneath this umbrella, we know that

27:15

it's going to. adhere to privacy, any

27:18

personal information will be removed so it

27:20

doesn't violate any problems. And then the

27:22

final piece is... Yeah we have the

27:24

tools but then we need that monthly

27:26

or bi-weekly lunch and learn where like

27:28

hey Scott what did you build this

27:31

week? Oh cool let's just duplicate that

27:33

one click and now send it to

27:35

Dan and Dan has similar work or

27:37

you know new employee starts they can

27:39

look over my shoulder it already the

27:41

bots already trained on all the history

27:43

it knows what to do so they

27:46

can jump in and you know I

27:48

always I always say that AI really

27:50

raises the floor you know like every

27:52

new employee could start at average. or

27:54

slightly above average. You still need to

27:56

raise the ceiling yourself, add that special

27:59

spice, right? Your own ideas, but it's

28:01

going to make everyone on a whole

28:03

quicker to get to work and higher,

28:05

I guess like higher average across the

28:07

board. And I always tell, you know,

28:09

the framework I always recommend is like

28:11

the AI sandwich. Like just think about

28:14

you, the AI interaction starts with you,

28:16

the human, the bread on top. Then

28:18

the AI is going to do something,

28:20

that's the meat in the middle. But

28:22

then you still have to be the

28:24

human on the bottom to take that

28:26

output and to improve it, to share

28:29

it, to repurpose it. And so I

28:31

think a lot of new people get

28:33

the bread and the meat, but they

28:35

forget the bottom piece of bread. And

28:37

so that'd be like the work I

28:39

would do as a leader is, here's

28:42

our tools, you can all use it,

28:44

and you're all going to be good.

28:46

Like you're not going to have spellingaling.

28:48

get better and it's going to be

28:50

like adding your own spice on that

28:52

last piece of bread. So that's what

28:54

I would do for Scott Inc. So

28:57

I think home run. And part of

28:59

that too is like developing the muscle

29:01

memory. So like for me, for example,

29:03

the, you know, we've been going through

29:05

fundraising recently, there's always like the same

29:07

set of questions that come up in

29:09

indiligence and in in in questions about

29:12

the product and all this and Most

29:14

of those have been answered like three

29:16

million times now in some form and

29:18

you know now looking back like and

29:20

you know we've started to do this

29:22

actually but really what would be best

29:25

is if we just had a little

29:27

chat that had all that preloaded into

29:29

it and could chat over that but

29:31

at the time it's like oh well

29:33

I'll just answer this email that's asking

29:35

these 10 questions right I can bang

29:37

that out really quick, but that, I

29:40

guess there's a muscle memory thing there,

29:42

and then there's a, there is some

29:44

barrier to overcome to configure the system

29:46

for future benefit, right, that you might

29:48

not see, see there, so. I don't

29:50

know, yeah, any, any suggestions even in

29:52

your own personal life where you've kind

29:55

of come over. I mean, we did

29:57

the same thing, right? Like, we did

29:59

a raise with Chip and we built

30:01

a chip chat and it was trained

30:03

on all of our, you know, slides

30:05

and everything. And people still want to

30:08

talk to you, like, it doesn't mean

30:10

that they don't get a human. Right,

30:12

exactly. But it gives them the option.

30:14

And like, you know, the data we're

30:16

seen for our users for our users

30:18

using chip for. for like, customer support,

30:20

like a chat bubble sort of use

30:23

case, 70% of them are not clicking

30:25

the talk to a human button. Like,

30:27

they just want to know what are

30:29

your opening hours, how much does it

30:31

cost, who are you, like, just give

30:33

me the facts. And as like a

30:36

busy parent, I get that, right? Like,

30:38

I don't want to make phone calls

30:40

because I know it'll take five to

30:42

10 minutes versus a minute if I'm

30:44

doing it myself. So, so I think

30:46

there's that aspect of like time efficiency.

30:48

And it is changing habits of like

30:51

going somewhere else or like you said

30:53

taking core info and putting it into

30:55

a repository What we found most helpful

30:57

is we have something called dynamic knowledge

30:59

sources So if it's a spreadsheet or

31:01

a folder on Google Drive or one

31:03

drive anything that gets added into those

31:06

places is automatically added into your agent.

31:08

And so I think with businesses, it's

31:10

important to think about that flow of

31:12

information and minimizing as much like documentation

31:14

work as you can. So we always

31:16

put everything into notion or confluence or

31:19

Google sheets or Google docs. Make that

31:21

your hub that is fed into the

31:23

AI. So everything that you put in

31:25

that place gets automatically added into your

31:27

FAQ bot or your marketing. assistant bot

31:29

or whatever. So I think that's that's

31:31

key is like you can ask people

31:34

to do it but even better is

31:36

like not to require more work or

31:38

even changing behavior because we know that's

31:40

the hardest part. So maybe it's a

31:42

VCC email that goes into a spreadsheet

31:44

that's automated right or you know something

31:46

like that so you can kind of

31:49

decide. The way we do it is

31:51

we actually look at our chat logs

31:53

of people engaging with CHIP and find

31:55

the answers that are going unanswered or

31:57

don't have a great answer. and then

31:59

we add those things in once a

32:02

week into our chat, so that it

32:04

improves for the things people are asking

32:06

for rather than trying to solve for

32:08

hypothetical edge cases. Yeah. Well, Scott, we've

32:10

kind of, we've talked a little bit

32:12

about CHIP, I've described it a little,

32:14

a little bit. I'm wondering maybe for,

32:17

you know, you've been on this journey

32:19

of kind of trying to build this

32:21

easy to use AI tool. along that

32:23

journey have you found, I'm sure you

32:25

tried various things that did work and

32:27

didn't work and certain things have been

32:29

difficult and certain things have been easier.

32:32

As you reflect on that kind of

32:34

as a founder of an AI company

32:36

trying to build an AI tool, any

32:38

things that you'd want to highlight in

32:40

terms of things that were kind of

32:42

key insights or bumps along the road

32:45

that in retrospect you look at and

32:47

kind of makes sense or anything like

32:49

that because I think there are a

32:51

lot in our audience that have maybe

32:53

ideas for things out there. Yeah. That's

32:55

amazing. There's so many. I think I'll

32:57

take like a non-obvious one which is

33:00

we've pretty early on focused on building

33:02

community so we have over 20 chip

33:04

chapters around the world people teaching one

33:06

another AI fairly active discord. That's been

33:08

invaluable because those are the people who

33:10

are bringing back. problems and ideas. And

33:13

being able to build towards actual customer

33:15

questions is so important. And a lot

33:17

of times customers don't have time or

33:19

interest in giving you feedback, which you

33:21

need. And so what we've done is

33:23

like. every two weeks or so basically

33:25

having free workshops to try to educate

33:28

our users and anybody and that's really

33:30

built a relationship I think where we

33:32

know these people by name we know

33:34

where they live what they do and

33:36

and it makes it a lot easier

33:38

for them to be like you know

33:40

can you build this thing I need

33:43

it for a pitch on Friday and

33:45

we're like yeah for you of course

33:47

because you're contributing you know so it's

33:49

building that building relationships and it doesn't

33:51

have to be hundreds right this can

33:53

be of people who love you. And

33:56

that's how you really start is like

33:58

a strong foundation. So I think that

34:00

one's not obvious. I think technically something

34:02

that we found maybe an accident and

34:04

we're trying to lean into now is

34:06

riding the wave of other people's innovation.

34:08

You know, like you can only build

34:11

so many unique pieces and you need

34:13

to be on top of other parts

34:15

of the tech stack. And so. chip

34:17

is built on top of large language

34:19

models. So as anthropic and open AI

34:21

build better models, chip gets better. And

34:23

for a lot of our users, they

34:26

think chip is doing that because, you

34:28

know, we are their front door to

34:30

AI. And so as the models get

34:32

better, chip gets better and their experience

34:34

gets better. We partner with folks like

34:36

prediction guard who help us provide better

34:39

privacy and security, right? And so. We

34:41

could go spend six months trying to

34:43

build that, but now we've lost the

34:45

whole point of what we're doing, right?

34:47

And so what is your forte is

34:49

really important. One thing that has really

34:51

recently that we kind of focused on

34:54

is Antropic has a new protocol called,

34:56

what is it, Model Context Protocol? It's

34:58

basically an easy way to connect APIs

35:00

in to AI tools. And so that's

35:02

another example of like we've been building.

35:04

one-off APIs to all these different tools.

35:07

And now it's like, wow, there's this

35:09

whole world that's built towards the standard.

35:11

And if we just tap into that,

35:13

now we can, again, get better, the

35:15

more the open source community contributes. So

35:17

I think that's really interesting to look

35:19

out where the areas that will move

35:22

quickly, that you can ride that wave,

35:24

and then where do you want to

35:26

be a differentiator? And you can kind

35:28

of draw your line wherever the right

35:30

places. you probably don't try to draw

35:32

it on all of them. Like pick

35:34

the ones you're best at. Yeah, I

35:37

think those are those are a few

35:39

and I think just the power of

35:41

small teams now. I mean, you read

35:43

that a lot of places, but you

35:45

know our CTO hunter who is just

35:47

like a beast with AI coding and

35:50

it's like, I know our output compared

35:52

to some legacy teams is just vastly

35:54

greater. And so I wouldn't underestimate if

35:56

you're a solo founder, you're a solo

35:58

founder, you're a in Colorado who he's

36:00

building a million dollar one person agency

36:02

and he's almost there right and it's

36:05

all built with AI automations and he's

36:07

conducting everything there's a lot of potential

36:09

out there so I would encourage you

36:11

when listening like finding a co-founder or

36:13

a team is really really hard but

36:15

you don't have to wait like you

36:17

can do a lot. on your own.

36:20

I'm curious, you actually started to get

36:22

in for a second to the next

36:24

question I was going to answer and

36:26

that was, you mentioned like privacy and

36:28

security and partnering with prediction guard for

36:30

that. As you're thinking about these

36:33

different concerns that weigh in on

36:35

various industries and you know there

36:37

will be legal concerns, things like

36:39

you know HIPAA in the medical

36:41

world and every industry has its

36:43

own set of concerns that are

36:45

kind of external but are binding

36:47

the work in those areas? And

36:49

as you are kind of unleashing

36:51

people's potential with the work that

36:53

you're doing, those kind of have

36:55

to find some sort of balance.

36:57

How are you thinking about the

36:59

constraints versus the unleashing that we

37:02

talked about and finding a balance

37:04

so that people are unleashed while

37:07

they're still having to be held

37:09

to account, you know, by whatever

37:11

those constraints in their industry is?

37:14

Right, yeah. I mean, I think

37:16

regulation's always going to trail the

37:18

innovation. And so I would say,

37:21

as a company, as an individual,

37:23

like, look at yourself first before

37:25

worrying about the regulatory environment. You

37:28

know, I think about privacy pyramid

37:30

as what we tell our customers, like

37:32

the bottom of the pyramid, the first

37:34

thing you should do is just think

37:36

about what are you okay sharing and

37:38

not sharing and just tell people. even

37:40

if it's hypothetical like not real like

37:42

I don't know that fear from elementary

37:44

school like sticks with us you know

37:46

and and so the first thing you

37:48

have to do is remove the fear

37:50

and the best way to do that is

37:52

just to say what the rules are as long

37:55

as people know the rules they'll work within them

37:57

but if they don't know what they are they're

37:59

afraid that whatever they do, we'll get them

38:01

in trouble, right? So, hey, just don't

38:03

upload customer data, like that's our rule.

38:05

Great, that's a great place to start.

38:07

Now go do anything else. Or, you

38:09

know, no customer data and don't integrate

38:11

with these files. Great. And the second

38:13

level of the pyramid after, you know,

38:15

kind of just best practices internally, is

38:18

then going to be like human protection

38:20

error, I call it, which, you know,

38:22

one thing prediction guard offers as well,

38:24

which is like. encrypting pieces of information

38:26

that I get added that shouldn't be

38:28

right so if I add a phone

38:30

number or you know a social security

38:32

number or something like it gets removed

38:34

for me because I made a mistake

38:36

that's fine like we make mistakes but

38:38

best practices, and then cover other people's

38:40

mistakes up as they make them. And

38:43

then I think the top of the

38:45

pyramid is where you actually say, you

38:47

know what, let's put it in our

38:49

own environment. So that way, if we

38:51

can share whatever we want without having

38:53

to worry, and that's where you can

38:55

run an open source, large language model

38:57

in your own cloud infrastructure, whatever you

38:59

share is in your cloud infrastructure. So,

39:01

you know. Some businesses have to do

39:03

that. So if you are in finance

39:05

health care, like you're probably going to

39:08

want to do that anyway just for

39:10

regulatory reasons. Some people want to do

39:12

that because they know they're going to

39:14

be sharing data that might be sensitive.

39:16

But I think for most of us

39:18

to get started, just follow that basic

39:20

best practice of like, think about it

39:22

before you share it. And if you're

39:24

working with a team that might make

39:26

mistakes or our contractors who aren't following

39:28

your rules, like add in that second

39:30

level of like human air protection. Scott,

39:33

as we kind of get near to

39:35

the end here, I'm wondering if you

39:37

can maybe share just a few standout

39:39

use cases of maybe things that you've

39:41

seen people do with CHIP that have

39:43

either surprised you or stood out in

39:45

a way like, oh, I didn't expect

39:47

people would do this, that, you know,

39:49

or things that are like, oh, I

39:51

didn't even know, you know, I built

39:53

the platform, but I didn't even know

39:56

that was possible. Every day, that's my

39:58

favorite part of chip and AI generally

40:00

is like we really are building the

40:02

tools and we don't know how people

40:04

will use them and it's so crazy

40:06

to see what people do with it

40:08

and I mean the most common use

40:10

cases I would say like there's kind

40:12

of five areas that people use all

40:14

the time. It's like operations marketing sales.

40:16

I call it company search like finding

40:18

stuff in your Google Drive basically and

40:21

What's the last one like data analysis

40:23

you like reviewing financials and things like

40:25

that so those are like the most

40:27

common But in terms of like fun

40:29

weird ones like we had somebody who

40:31

launched a Canadian tariff checker and so

40:33

like as the tariffs on Canada were

40:35

released you could actually search any product

40:37

and it would source like where they

40:39

were coming from and tell you what

40:41

the change in price would be that

40:43

was like totally interesting One of my

40:46

favorite use cases is a guy named

40:48

Tyler Hansen. He's in Sioux Falls, South

40:50

Dakota, and he runs an HVAC company.

40:52

And he put in all of the

40:54

training manuals for all of the equipment

40:56

that they service. So then his technicians

40:58

are on the ground. And instead of

41:00

having to like. be in the bathroom

41:02

watching a YouTube video, which I know

41:04

has happened when my HVAC guy comes,

41:06

right? He's like actually learning how to

41:08

do the thing that I asked him

41:11

to do. Like they can actually pull

41:13

up the specific model via their chip

41:15

chat and get instructions on what to

41:17

do and how to service it and

41:19

parts and that one's really fun. There's

41:21

a contractor out in Washington. He uses

41:23

it to create supply list. So he

41:25

just puts in square footage and what

41:27

people are going to build and what

41:29

people are going about. A lot of

41:31

people doing it for like finding HR

41:33

policies, finding, let's see, there's a car

41:36

dealer that's using it to find cars

41:38

to purchase, like to then resell, right?

41:40

So like searches through Auto Trader and,

41:42

you know, Craigslist or wherever else to

41:44

find vehicles. Just so many things, right?

41:46

Every day I'm encountering new ones that

41:48

are so fascinating. The fun part is

41:50

we integrate with you know, APIs and

41:52

web hooks. So really like any tool

41:54

can get pulled in and a lot

41:56

of times chip ends up being a

41:58

front end to an AI tool that's

42:01

talking to their software. So chip becomes

42:03

a way they communicate, but then it's

42:05

pulling their own data. So that's super

42:07

fun. Personally, I have a Scott bot,

42:09

you know, that's the one I use

42:11

every single day. And so, like, I

42:13

can write things very quickly and remember

42:15

people that I've talked to, so I

42:17

can, like, brings in past conversations. And

42:19

so that helps me quite a bit.

42:21

So yeah, those are a few. random

42:23

ideas. I haven't built the West Lafayette

42:26

tour guide yet, but we do have

42:28

some travel, travel AI tools out there,

42:30

so I bet we could do that

42:32

too. So very cool. And while you're,

42:34

while you're building that tour guide, I

42:36

might give you a location or two

42:38

as well. Okay, there you go. Yeah,

42:40

that's awesome. So really cool use cases

42:42

there as you, like, that's got to

42:44

get you thinking about like the possibilities.

42:46

So, you know, you come at it

42:48

with your own mindset. your customers are

42:51

teaching you every day about what the

42:53

new possibilities and boundaries might be. So

42:55

where does that take you? Like when

42:57

you are, you know, you're kind of

42:59

done for the workday, your brain's decompressing,

43:01

but you're still kind of, you know,

43:03

just working on things, what's going through

43:05

your head about, like, where could things

43:07

go with this? You know, you take.

43:09

what you're driving and the folks you're

43:11

working with are driving you're taking what

43:14

your customers are showing you that you

43:16

never thought about and that's gonna leave

43:18

you with some pretty cool ideas about

43:20

what the future might hold but can

43:22

you share some of those ideas with

43:24

us? Yeah I think I mean I

43:26

reflect at the end of the day

43:28

in a lot of ways because I

43:30

have four kids that are 11 973

43:32

and I just really try to think

43:34

about like what does society look like

43:36

when this is more present and you

43:39

know, what does education look like? I

43:41

spent a lot of my life in

43:43

education. We work with a lot of

43:45

schools who use it for tutors and

43:47

advisors and, you know, what's the value

43:49

of a credential saying, you know, something

43:51

when the pace of change is like

43:53

way faster than four years, right? I

43:55

think ultimately, you know, I imagine this

43:57

technology has to fade away from being

43:59

AI and just being apart. part of

44:01

what we use. And it helps us

44:04

lean into the things that make us

44:06

weird. You know, I think about AI

44:08

as the world's best cover band, and

44:10

it needs like the originals to cover.

44:12

And so I think it really forces

44:14

us to be more unique as individuals

44:16

and create something new. We're gonna use

44:18

AI for a lot of the quick

44:20

answers and it's gonna be average. It's

44:22

gonna be the middle of that bell

44:24

curve and that'll be fine for most

44:26

work. But again, we have to raise

44:29

the ceiling ourselves. And so. I think

44:31

it makes me feel like I want

44:33

my kids and hopefully myself to like

44:35

just get good, really good at whatever

44:37

weird interesting thing we care about. Yeah,

44:39

and I, man, I don't know, I

44:41

think agency again, like I keep coming

44:43

back to that, but how do you

44:45

instil a lack, like a fearlessness in

44:47

people? Because it feels like, first of

44:49

all, most people aren't aware of the

44:51

pace of change, and as they become

44:54

aware of it, it's either. I'm scared,

44:56

I'm going to back away, or I'm

44:58

going to lean into it. And I

45:00

think we just really need to lean

45:02

into it. And I don't know, I

45:04

think it's exciting because I'm in Fargo

45:06

and I couldn't, you know, learn to

45:08

be a nuclear physicist in Fargo, right?

45:10

But now I could. Like I can

45:12

easily go down that path and learn

45:14

what I need to connect with the

45:16

resources, you know, showcase my work. And

45:19

this has kind of been my dream

45:21

since my first company in 2009 of

45:23

like. really giving anyone wherever they are

45:25

a chance to build. And AI is

45:27

just like the next step in that

45:29

process. And I know a lot of

45:31

people still will find reasons not to,

45:33

but it's going to be just on

45:35

that agency piece, like you can. So

45:37

I don't know. I think a society

45:39

where everybody has a chance to build

45:41

and create is incredibly exciting. It's going

45:44

to be more competitive. Everyone around the

45:46

world has equal access to the same

45:48

models as NASA and, you know, like

45:50

the Defense Department. Like it's kind of

45:52

wild that you can log into these

45:54

things for free and have the same

45:56

power as everyone else. So that's an

45:58

opportunity if you if you take it.

46:00

I think I saw there's a recent

46:02

study the world banked it in Nigeria.

46:04

and students who are using chat GPT

46:07

as a tutor for six weeks had

46:09

the equivalent of two years of education.

46:11

And it's just so many of our

46:13

problems are problems of access, and I

46:15

think a lot of those access problems

46:17

go away. And then what happens when

46:19

another, you know, one billion people come

46:21

online with education who don't have it

46:23

now? Like, that's just better for us

46:25

all. We can come up with really

46:27

exciting solutions to our problems. Well said,

46:29

yeah, that's a that's a great way

46:32

to end. Thanks for, thanks for joining

46:34

Scott. I encourage everyone to go create

46:36

your first chip chat on chip chi-p-p-p.a.i.

46:38

and have some fun. Explore those, that

46:40

weirdness as Scott put it. I love

46:42

that. Thanks for joining Scott. It's been

46:44

great to be here, guys. All

46:52

right, that is our show

46:54

for this week. If you

46:56

haven't checked out our change

46:58

log newsletter, head to change

47:00

log.com/news. There you'll find 29

47:02

reasons, yes, 29 reasons why

47:04

you should subscribe. I'll tell

47:06

you reason number 17, you

47:08

might actually start looking forward

47:10

to Mondays. Sounds like somebody's

47:12

got a case of the

47:14

Mondays. 28 more reasons are

47:16

waiting for you at change

47:18

log.com/news. Thanks again to our

47:20

partners at fly.io to break

47:22

master cylinder for the beats

47:24

and to you for listening.

47:26

That is all for now,

47:28

but we'll talk to you

47:30

again next time.